CONF Imseng_INTERSPEECH_2011/IDIAP Improving non-native ASR through stochastic multilingual phoneme space transformations Imseng, David Bourlard, Hervé Dines, John Garner, Philip N. Magimai-Doss, Mathew EXTERNAL https://publications.idiap.ch/attachments/papers/2011/Imseng_INTERSPEECH_2011.pdf PUBLIC https://publications.idiap.ch/index.php/publications/showcite/Imseng_Idiap-RR-19-2011 Related documents Proceedings of Interspeech Florence, Italy 2011 537-540 We propose a stochastic phoneme space transformation technique that allows the conversion of conditional source phoneme posterior probabilities (conditioned on the acoustics) into target phoneme posterior probabilities. The source and target phonemes can be in any language and phoneme format such as the International Phonetic Alphabet. The novel technique makes use of a Kullback-Leibler divergence based hidden Markov model and can be applied to non-native and accented speech recognition or used to adapt systems to underresourced languages. In this paper, and in the context of hybrid HMM/MLP recognizers, we successfully apply the proposed approach to non-native English speech recognition on the HIWIRE dataset. REPORT Imseng_Idiap-RR-19-2011/IDIAP Improving non-native ASR through stochastic multilingual phoneme space transformations Imseng, David Bourlard, Hervé Dines, John Garner, Philip N. Magimai-Doss, Mathew multilingual acoustic modeling universal phoneme set EXTERNAL https://publications.idiap.ch/attachments/reports/2011/Imseng_Idiap-RR-19-2011.pdf PUBLIC Idiap-RR-19-2011 2011 Idiap June 2011 We propose a stochastic phoneme space transformation technique that allows the conversion of conditional source phoneme posterior probabilities (conditioned on the acoustics) into target phoneme posterior probabilities. The source and target phonemes can be in any language and phoneme format such as the International Phonetic Alphabet. The novel technique makes use of a Kullback-Leibler divergence based hidden Markov model and can be applied to non-native and accented speech recognition or used to adapt systems to under-resourced languages. In this paper, and in the context of hybrid HMM/MLP recognizers, we successfully apply the proposed approach to non-native English speech recognition on the HIWIRE dataset.